Editing text and images in place
Double-clicking a heading, paragraph or rich-text block edits it in place; an image swaps through the properties rail — nothing to type over a picture.
Double-clicking a heading, a paragraph, or a block of rich text turns that exact element into what you're typing into — the real rendered element, in the page's own type and spacing, not a form field swapped in over it.
The same save path as the properties rail
Committing an inline edit writes the identical property the properties rail would, through the identical action, so undo and redo behave exactly as they do for any other edit. There's nothing separate to keep track of.
Which piece of a module you're editing
Inside a module with more than one editable piece — a card's heading and its body, say — double-clicking lands on whichever one you clicked, not the module as a whole. A module with only one editable piece accepts a double-click anywhere on it. A double-click on the editor's own controls, or somewhere genuinely ambiguous between two editable pieces, edits nothing.
What doesn't work this way
A handful of modules hold something other than the text they render: html and embed hold raw source, tabs, carousel, tagCloud and breadcrumbs hold structured configuration, spacer holds a CSS length, and feedDisplay / categoryList hold a slug. Double-clicking any of these opens the properties rail instead of turning something contenteditable — there's nothing on the page itself that would be safe to type over.
An image module's first property is the picture, not a caption, so double-clicking one does the same thing: it opens the properties rail's picker rather than putting a cursor on the image. Swap the picture there.